Rspamd itself does not alter message in any way. However, it keeps all urls with their relative positions during scan. Moreover, it can resolve redirects for urls (e.g. for t.co or goo.gl) and provide 'real' links. But the second part of job: links analysis and rewriting is not provided by rspamd.
Well, I've recently ported a massive setup of SA rules and I've tried to add all functions that are supported by SA to rspamd. The only ones unsupported are pyzor/razor/dcc (all that is covered by fuzzy check: https://rspamd.com/doc/modules/fuzzy_check.html).
Both systems support statistics (rspamd uses 5-gramms hidden Markov model and SA uses naive bayes), they both support different backends for statistics (redis and sqlite for rspamd), per-user statistics, autolearning. Among network checks, they support URIBLs, RBLs, DMARC, DKIM, SPF. Obviously, both systems support regexp rules. However, I'm not an expert in SA features and might thus miss something...
In brief, rspamd can use the most of SA rules but it provides more optimization tricks than SA does. However, it is less mature and could be harder to setup than SA. You can also check my recent FOSDEM talk: https://www.youtube.com/watch?v=_fl9i-az_Q0
I'm really sorry that this information is not in the landing page of rspamd.com - that's obviously my fault.
From what you are saying, I can conclude that you are using very high scoring for statistical classifier (or basing solely on statistics). This is not an option for a system with millions of users (their accept/reject rate is close to 70/30 percents, as we cannot rely on bayes at all). Therefore, I've never ever evaluated bayes as a single classifier. Nevertheless, I'm using OSB-Bayes as a statistical algorithm for rspamd which has been proven to be a good classifier.
I've compared rspamd on random stream of user's messages with SA and Kaspersky antispam several years ago. And I've got almost the same rate of false positives and false negatives for all three products. However, over years spammers are getting much smarter (images spam, valid DKIM, valid SPF and other clever tricks).
Regarding statistics, rspamd uses OSBF-Bayes classifier and 5-gramms input (so it is not naive bayes). I've used the following academic paper: http://osbf-lua.luaforge.net/papers/osbf-eddc.pdf as reference. This algorithm is also used for crm114 spam classifier. However, bayes classifier is a very small part of rspamd (unlike dspamd, for example) and it could be almost useless if you have, let's say, 50 millions of users accounts. Rspamd is targeted for this grade systems.
They are not related at all. From what I'm observing, mopher is a milter (so it might be compared with my another project called 'rmilter') and it can work with Spamassassin. Hence, if something can work with spamassassin then it will be capable to talk with rspamd at some extent.
Multiple database drivers are in plans for rspamd 1.0 (along with personal statistics and advanced rules planner). The tricky stuff here is that rspamd uses non-blocking model currently which is hardly supported by database drivers (excluding redis and some others). However, rspamd has a concept of asynchronous threads executed in thread pool. So something like MySQL query could be executed within this thread pool with no delay for other filters processing.
This information is merely interesting for some academic researches on some very specific messages corpus. But in the real world, I cannot efficiently evaluate the accuracy because it depends on zillions of parameters. Moreover, since rspamd uses not only statistics but a number of sources, such as DNS lists, SPF, DKIM, hashes databases and so on, it is literally impossible to be determined about preciseness.